Deep learning-based method for non-uniform motion-induced error reduction in dynamic microscopic 3D shape measurement

Abstract

The non-uniform motion-induced error reduction in dynamic fringe projection profilometry is complex and challenging. Recently, deep learning (DL) has been successfully applied to many complex optical problems with strong nonlinearity and exhibits excellent performance. Inspired by this, a deep learning-based method is developed for non-uniform motion-induced error reduction by taking advantage of the powerful ability of nonlinear fitting. First, a specially designed dataset of motion-induced error reduction is generated for network training by incorporating complex nonlinearity. Then, the corresponding DL-based architecture is proposed and it contains two parts: in the first part, a fringe compensation module is developed as network pre-processing to reduce the phase error caused by fringe discontinuity; in the second part, a deep neural network is employed to extract the high-level features of error distribution and establish a pixel-wise hidden nonlinear mapping between the phase with motion-induced error and the ideal one. Both simulations and real experiments demonstrate the feasibility of the proposed method in dynamic macroscopic measurement.

Method for measuring the three-dimensional profile of moving objects based on neural network position transformation

“…By recombining the calculated results, six images with different phases at the same position were obtained. The images were reconstructed using a six-step phase-shifting method and compared with the RMER 14 and DNER 34 methods. The results are shown in Fig.…”

“…Yu et al 33 used deep learning phase retrieval techniques for dynamic 3D measurement. Tan et al 34 proposed a method using CNN to eliminate errors in phase unwrapping in the phase-shifting method. However, for the 3D measurement of moving objects, previous studies mostly used non-CNN methods to eliminate motion-induced errors, but these methods often have limitations on the object itself or the motion scene, and some researchers have used CNNs to eliminate motion-induced errors, but they only processed a single image with high noise levels.…”

“…used deep learning phase retrieval techniques for dynamic 3D measurement. Tan et al 34 . proposed a method using CNN to eliminate errors in phase unwrapping in the phase-shifting method.…”

Weakly Supervised Depth Estimation for 3D Imaging with Single Camera Fringe Projection Profilometry

“…However, the use of multiple images also dramatically limits the speed [ 5 ]. In addition, the phase-shifting method assumes that the object to be measured remains stationary during each 3D imaging so that motion artifacts will affect the 3D imaging accuracy [ 6 , 7 ].…”

Large depth range dithered binary focusing fringe projection technique

“…measurement accuracy. Recently, deep learning has been successfully applied to fringe analysis [11,12] Inspired by this, we develop a dithered binary fringe pattern dataset and design a generative adversarial network for fringe pattern translation. The raw binary patterns are obtained by focusing projection and then the sinusoidal patterns can be generated through network optimization.…”